*** merbot <merbot!~supybot@phost1.merproject.org> has joined #sailfishos-push | 09:13 | |
*** abranson_ <abranson_!~abranson@greenback.default.abranson.uk0.bigv.io> has joined #sailfishos-push | 14:02 | |
abranson_ | hello, is this still on? | 14:04 |
---|---|---|
abranson_ | tbr: Florian Leeber is asking if anyone would be interested in a joint SFOS/UBPorts open push notification service. | 14:06 |
tbr | abranson_: heh, canonical already killing their infra? now that's fast | 14:09 |
tbr | abranson_: I'm happy to share my thoughts and test results with them | 14:12 |
abranson_ | tbr: yes it does seem a little quick. that's the problem with such tethered services though eh? | 14:15 |
abranson_ | thanks, I'll let him know. probably best if he comes in here | 14:15 |
tbr | ayup | 14:16 |
tbr | I think the canonical stuff is mostly open sauce | 14:16 |
tbr | so we could look into reusing that or most of it | 14:16 |
abranson_ | he said it was part of the ubuntu one thing, so that might make it a bit less likely | 14:19 |
tbr | well, I looked into it back then when they evaluated mosquitto and then went full out NIH, and it looked like most the source was there. I guess account infrastructure and such might be an issue though. | 14:22 |
abranson_ | They went NIH again? what a shocker ;) | 14:26 |
tbr | yeah, *nobody* could have seen that coming | 14:27 |
*** Flohack <Flohack!~Flohack@91.118.126.194> has joined #sailfishos-push | 14:33 | |
Flohack | Hi there | 14:34 |
Flohack | Florian from UBports | 14:34 |
Flohack | The idea came up to work on a global unified open-source push platform ;) | 14:34 |
tbr | lo | 14:36 |
tbr | so, canonical already pulling the plug on their NIH push thing? | 14:36 |
tbr | I did some work previously for sailfish, as that was then lacking such a service (and still is AFAIU). | 14:38 |
tbr | Did a proof of concept based on MQTT (mosquitto), which worked very reliably | 14:38 |
Flohack | Yeah Canonical will kill things slowly but steady. Since a few weeks users reporting more and more issues with notifications failing | 14:39 |
Flohack | I read the page already yes | 14:39 |
tbr | you probably know better than me. How much of the ubuntu touch notification infrastructure is open source? | 14:40 |
tbr | I stopped looking at it pretty early on, when they kept playing hide and seek with me. | 14:40 |
Flohack | Nothing is open source, they just want to give us an example/demoe code that works "similar" | 14:42 |
Flohack | ^^ | 14:42 |
Flohack | So we decided not to pursue this idea any longer | 14:42 |
Flohack | to cooperate with Canonical, its not possible | 14:42 |
tbr | ugh | 14:44 |
tbr | worse than I expected :-/ | 14:44 |
Flohack | It gets worse for us at least: Telegram sends now all notifications to Canonical server. Rumors have it that Canonical actually has to pay for that :) => So how can we convince Telegram to send us all their stuff for free xD | 14:46 |
tbr | ok, so there to be honest, you should prepare yourselves for loss of service. | 14:57 |
Flohack | Yes we know | 15:01 |
Flohack | There are 100 ideas how to work around, but in fact we cant go the way to say, ok then phone checks every 5 minutes on Telegram, since that means a complete login to the cloud, iterating through lots of lists to finally know if smth new is there or not | 15:02 |
Flohack | For POP3, IMAP it would be less complicated | 15:02 |
tbr | most IMAP servers speak IDLE by now, that's essentially push. I'm not sure how many client libraries have that implemented. | 15:14 |
Flohack | But then again the socket must stay open, the app must stay active | 15:17 |
tbr | yes, just like in other push protocols | 15:18 |
Flohack | This can be discussed in detail but currently the idea is that the Email app does not run all the checks by itself, but has a little companion plugin that is being called by the push helper to do its job. So it would login/logout to check | 15:21 |
tbr | anyway, so what would be the aim? build a push protocol stack from scratch? mimic the ubuntu push server? | 15:25 |
Flohack | The first step is that we find a list of requirements that it should fulfill | 15:35 |
Flohack | maybe we can use your message queueing on the lower level, could be better than a rewrite | 15:35 |
Flohack | I can open a wiki page to start brainstorming if you like | 15:38 |
tbr | sure | 15:42 |
tbr | if people don't mind we can use this channel for now | 15:42 |
tbr | I'm open to discuss basic architecture and concepts | 15:42 |
abranson_ | I've added a comment to a relevant bug in the internal Jolla bz so people know about this. I don't think it would be a priority by any stretch, but there has been interest in it in the past. Especially while there are no daemons allowed in the harbour. | 15:42 |
tbr | I have a test server up that we can use to evaluate/iterate, etc | 15:43 |
Flohack | Oki | 15:43 |
tbr | the concept was always meant to be platform independent, it just happened to be that at that time it scratched my itch of push notifications on sailfishos | 15:44 |
Flohack | Yeah we can use this channel but I like to persist things in one place. So basically I can write summaries from here to https://wiki.ubports.com/wiki/Universal-Push-Notifications%3A-WIP%2C-Brainstorming%2C-etc. | 15:47 |
Flohack | tbr: How you plan to authenticate users/devices? | 15:49 |
Flohack | On my way home now. I will return here, but I do not have an IRC logger online ;) | 15:59 |
tbr | sure | 16:09 |
tbr | http://www.merproject.org/logs/%23sailfishos-push/ - I should turn on the logg prettyfy thing for this | 16:12 |
*** Flohack <Flohack!~Flohack@91.118.126.194> has quit IRC (Remote host closed the connection) | 16:14 | |
*** Flohack <Flohack!~flohack@2a02:8388:8981:d00:243f:7cd1:1e7a:cf8f> has joined #sailfishos-push | 17:25 | |
Flohack | Re | 17:25 |
Flohack | Isnt MQTT the thing that is used by Facebook? | 17:26 |
tbr | IIRC they use that for *something* but not sure if it's in their push protocol itself | 17:28 |
tbr | 16:12:11< tbr> http://www.merproject.org/logs/%23sailfishos-push/ - I should turn on the logg prettyfy thing for this | 17:28 |
tbr | in case you need logs | 17:29 |
Flohack | So basically I read about it, there are possibilities to form a cluster of MQTT brokers. That would be a good thing | 17:29 |
Flohack | Thx | 17:29 |
tbr | mosquitto scales massively by itself already, but yes you can cluster it | 17:29 |
Flohack | Yeah its jusst for availability, I would put there lets say 3 nodes for the beginning. As I see it could support 1000s of devices | 17:30 |
Flohack | So we would be good and anyway we would monitor system load | 17:30 |
tbr | I'd need to go back to my mosquitto ml archives, but I think someone loaded it with tens of thousands of connections and massive throughput | 17:32 |
tbr | btw: MQTT doesn't play well with proxies | 17:32 |
Flohack | You mean people behind NAT? | 17:33 |
tbr | no | 17:33 |
tbr | NAT is a separate problem, there you just need to send keepalives to keep the NAT table entries fresh | 17:33 |
tbr | I never had any NAT problems with my test service | 17:34 |
tbr | especially enterprise networks might block any outbound traffic unless it goes through a HTTP proxy | 17:34 |
Flohack | Hmm ok but then in the end the user will find out about this. Must be the same for Android/iOS | 17:35 |
tbr | MQTT will not always work in such settings. It *might*. | 17:35 |
tbr | yes, although I think Android and iOS do funky HTTP tunneling fallbacks | 17:36 |
tbr | something that could of course be also implemented in the long run | 17:36 |
Flohack | sure | 17:37 |
Flohack | Hmm funny the whole push client on Ubuntu Touch is written in Go | 17:38 |
Flohack | I cant understand this xD | 17:38 |
*** Flohack <Flohack!~flohack@2a02:8388:8981:d00:243f:7cd1:1e7a:cf8f> has quit IRC (Quit: ChatZilla 0.9.93 [Firefox 53.0.3/20170518000419]) | 21:30 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!